Re: Does TCP Need an Overhaul? (internetevolution, via slashdot)
Kevin Day wrote: Yeah, I guess the point I was trying to make is that once you throw SACK into the equation you lose the assumption that if you drop TCP packets, TCP slows down. Before New Reno, fast-retransmit and SACK this was true and very easy to model. Now you can drop a considerable number of packets and TCP doesn't slow down very much, if at all. If you're worried about data that your clients are downloading you're either throwing away data from the server (which is wasting bandwidth getting all the way to you) or throwing away your clients' ACKs. Lost ACKs do almost nothing to slow down TCP unless you've thrown them *all* away. If this was true surely it would mean that drop models such WRED/RED are becoming useless? Sam
Re: Does TCP Need an Overhaul? (internetevolution, via slashdot)
On 5 apr 2008, at 12:34, Kevin Day wrote: As long as you didn't drop more packets than SACK could handle (generally 2 packets in-flight) dropping packets is pretty ineffective at causing TCP to slow down. It shouldn't be. TCP hovers around the maximum bandwidth that a path will allow (if the underlying buffers are large enough). It increases its congestion window in congestion avoidance until a packet is dropped, then the congestion window shrinks but it also starts growing again. If you read The macroscopic behavior of the TCP Congestion Avoidance algorithm by Mathis et al you'll see that TCP performance conforms to: bandwdith = MSS / RTT * C / sqrt(p) Where MSS is the maximum segment size, RTT the round trip time, C a constant close to 1 and p the packet loss probability. Since the overshooting of the congestion window causes congestion = packet loss, you end up at some equilibrium of bandwidth and packet loss. Or, for a given link: number of flows, bandwidth and packet loss. I'm sure this behavior isn't any different in the presence of SACK. However, the caveat is that the congestion window never shrinks between two maximum segment sizes. If packet loss is such that you reach that size, then more packet loss will not slow down sessions. Note that for short RTTs you can still move a fair amount of data in this state, but any lost packet means a retransmission timeout, which stalls the session. You've also got fast retransmit, New Reno, BIC/CUBIC, as well as host parameter caching to limit the affect of packet loss on recovery time. The really interesting one is TCP Vegas, which doesn't need packet loss to slow down. But Vegas is a bit less aggressive than Reno (which is what's widely deployed) or New Reno (which is also deployed but not so widely). This is a disincentive for users to deploy it, but it would be good for service providers. Additional benefit is that you don't need to keep huge numbers of buffers in your routers and switches because Vegas flows tend to not overshoot the maximum available bandwidth of the path.
Re: Does TCP Need an Overhaul? (internetevolution, via slashdot)
On Apr 7, 2008, at 7:17 AM, Iljitsch van Beijnum wrote: On 5 apr 2008, at 12:34, Kevin Day wrote: As long as you didn't drop more packets than SACK could handle (generally 2 packets in-flight) dropping packets is pretty ineffective at causing TCP to slow down. It shouldn't be. TCP hovers around the maximum bandwidth that a path will allow (if the underlying buffers are large enough). It increases its congestion window in congestion avoidance until a packet is dropped, then the congestion window shrinks but it also starts growing again. I'm sure this behavior isn't any different in the presence of SACK. At least in FreeBSD, packet loss handled by SACK recovery changes the congestion window behavior. During a SACK recovery, the congestion window is clamped down to allow no more than 2 additional segments in flight, but that only lasts until the recovery is complete and quickly recovers. (That's significantly glossing over a lot of details that probably only matter to those who already know them - don't shoot me for that not being 100% accurate :) ) I don't believe that Linux or Windows are quite that aggressive with SACK recovery though, but I'm less familiar there. As a quick example on two FreeBSD 7.0 boxes attached directly over GigE, with New Reno, fast retransmit/recovery, and 256K window sizes, with an intermediary router simulating packet loss. A single HTTP TCP session going from a server to client. SACK enabled, 0% packet loss: 780Mbps SACK disabled, 0% packet loss: 780Mbps SACK enabled, 0.005% packet loss: 734Mbps SACK disabled, 0.005% packet loss: 144Mbps (19.6% the speed of having SACK enabled) SACK enabled, 0.01% packet loss: 664Mbps SACK disabled, 0.01% packet loss: 88Mbps (13.3%) However, this falls apart pretty fast when the packet loss is high enough that SACK doesn't spend enough time outside the recovery phase. It's still much better than without SACK though: SACK enabled, 0.1% packet loss: 48Mbps SACK disabled, 0.1% packet loss: 36Mbps (75%) However, the caveat is that the congestion window never shrinks between two maximum segment sizes. If packet loss is such that you reach that size, then more packet loss will not slow down sessions. Note that for short RTTs you can still move a fair amount of data in this state, but any lost packet means a retransmission timeout, which stalls the session. True, a longer RTT changes this effect. Same test, but instead of back- to-back GigE, this is going over a real-world trans-atlantic link: SACK enabled, 0% packet loss: 2.22Mbps SACK disabled, 0% packet loss: 2.23Mbps SACK enabled, 0.005% packet loss: 2.03Mbps SACK disabled, 0.005% packet loss: 1.95Mbps (96%) SACK enabled, 0.01% packet loss: 2.01Mbps SACK disabled, 0.01% packet loss: 1.94Mbps (96%) SACK enabled, 0.1% packet loss: 1.93Mbps SACK disabled, 0.1% packet loss: 0.85Mbps (44%) (No, this wasn't a scientifically valid test there, but the best I can do for an early Monday morning) You've also got fast retransmit, New Reno, BIC/CUBIC, as well as host parameter caching to limit the affect of packet loss on recovery time. The really interesting one is TCP Vegas, which doesn't need packet loss to slow down. But Vegas is a bit less aggressive than Reno (which is what's widely deployed) or New Reno (which is also deployed but not so widely). This is a disincentive for users to deploy it, but it would be good for service providers. Additional benefit is that you don't need to keep huge numbers of buffers in your routers and switches because Vegas flows tend to not overshoot the maximum available bandwidth of the path. It would be very nice if more network-friendly protocols were in use, but with download optimizers for Windows that cranks the TCP window sizes way up, the general move to solving latency by opening more sockets, and P2P doing whatever it can to evade ISP detection - it's probably a bit late. -- Kevin
Superfast internet may replace world wide web
says the solemn headline of Telegraph. http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/06/ninternet106.xml Also related to this one, here: Web could collapse as video demand soars http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/07/nweb107.xml .. and we in Nanog are still discussing IPv6! ;-)
Re: Does TCP Need an Overhaul? (internetevolution, via slashdot)
Thanks for sharing your test results Kevin. Most interesting. I am setting up a small test lab of a couple linux boxes myself to learn more about the various traffic shaping and TCP stack options. Ill post my results here. I am primarily interested in local wireless network performance optimization vs long haul connects at least for now. It would be interesting to see how the numbers change with a windows or linux box on one end and BSD on the other. Also how did you simulate packet loss? Sent via BlackBerry from T-Mobile
Re: Superfast internet may replace world wide web
On Mon, 7 Apr 2008, Bill Woodcock wrote: On Mon, 7 Apr 2008, Glen Kent wrote: says the solemn headline of Telegraph. .. and we in Nanog are still discussing IPv6! ;-) It's because we don't have a hadron demolition derby to power our American interwebs: The power of the grid will be unlocked this summer with the switching on of the Large Hadron Collider (LHC). http://xkcd.com/401/ -Bill
Re: Superfast internet may replace world wide web
On Mon, 7 Apr 2008, Glen Kent wrote: says the solemn headline of Telegraph. .. and we in Nanog are still discussing IPv6! ;-) It's because we don't have a hadron demolition derby to power our American interwebs: The power of the grid will be unlocked this summer with the switching on of the Large Hadron Collider (LHC). -Bill
Re: Superfast internet may replace world wide web
Glen Kent wrote: says the solemn headline of Telegraph. http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/06/ninternet106.xml It is always good to see that journalists don't know that Networks are also used for other purposes than their daily dose of nonsense (also called the Internet or World Wide Web for the web-only portion etc) Also related to this one, here: Web could collapse as video demand soars http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/07/nweb107.xml .. and we in Nanog are still discussing IPv6! ;-) The CERN LHC (Large Hadron Collider) is a nice toy. They had an Open Day (http://lhc2008.web.cern.ch/LHC2008/index-E.html) yesterday, which was really impressive. No more Open Days in the tunnel are planned for the next couple of years, thus for everybody who missed it: http://gallery.unfix.org/2008/2008-04-06-cern-lhc/ Greets, Jeroen signature.asc Description: OpenPGP digital signature
Re: Superfast internet may replace world wide web
Bill Woodcock wrote: On Mon, 7 Apr 2008, Glen Kent wrote: says the solemn headline of Telegraph. .. and we in Nanog are still discussing IPv6! ;-) It's because we don't have a hadron demolition derby to power our American interwebs: The power of the grid will be unlocked this summer with the switching on of the Large Hadron Collider (LHC). And those of us that live next to the LHC wonder if we will be sucked into a {vortex|wormhole}. Thomas
Re: Does TCP Need an Overhaul? (internetevolution, via slashdot)
On Mon, 7 Apr 2008, [EMAIL PROTECTED] wrote: Thanks for sharing your test results Kevin. Most interesting. I am setting up a small test lab of a couple linux boxes myself to learn more about the various traffic shaping and TCP stack options. Ill post my results here. I am primarily interested in local wireless network performance optimization vs long haul connects at least for now. Anyone out there attend this event? The Future of TCP: Train-wreck or Evolution? http://yuba.stanford.edu/trainwreck/agenda.html how did the demos go? - Lucy It would be interesting to see how the numbers change with a windows or linux box on one end and BSD on the other. Also how did you simulate packet loss? Sent via BlackBerry from T-Mobile
Re: Superfast internet may replace world wide web
On Mon, 7 Apr 2008 08:24:54 -0700 (PDT) Lucy Lynch [EMAIL PROTECTED] wrote: On Mon, 7 Apr 2008, Bill Woodcock wrote: On Mon, 7 Apr 2008, Glen Kent wrote: says the solemn headline of Telegraph. .. and we in Nanog are still discussing IPv6! ;-) It's because we don't have a hadron demolition derby to power our American interwebs: The power of the grid will be unlocked this summer with the switching on of the Large Hadron Collider (LHC). http://xkcd.com/401/ Also http://ars.userfriendly.org/cartoons/?id=20080330 and http://ars.userfriendly.org/cartoons/?id=20080406 --Steve Bellovin, http://www.cs.columbia.edu/~smb
Any tool or theorical method on detecting number of computer behind a NAT box?
hi, Sharing internet access bandwidth between multiple computers is common today. Usually, bandwidth sharer bought a little router with NAT/PAT function. After connecting that box to a ADSL/LAN access link, multiple computer could share a single access link. I heard some company provide prdouct for detecting number of computers behind NAT/PAT box. Is there any paper or document on how such product work? where could I fint them ? Joe __ Search, browse and book your hotels and flights through Yahoo! Travel. http://sg.travel.yahoo.com
Re: Superfast internet may replace world wide web
On Mon, 7 Apr 2008, Glen Kent wrote: says the solemn headline of Telegraph. http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/06/ninternet106.xml So, the internet was created in Switzerland at Cern's particle physics center? Can someone look up Al Gore's passport history and tell us when he spent time there? :) -- Jon Lewis | I route Senior Network Engineer | therefore you are Atlantic Net| _ http://www.lewis.org/~jlewis/pgp for PGP public key_
Re: Superfast internet may replace world wide web
On Apr 7, 2008, at 11:36 AM, Thomas Kernen wrote: Bill Woodcock wrote: On Mon, 7 Apr 2008, Glen Kent wrote: says the solemn headline of Telegraph. .. and we in Nanog are still discussing IPv6! ;-) It's because we don't have a hadron demolition derby to power our American interwebs: The power of the grid will be unlocked this summer with the switching on of the Large Hadron Collider (LHC). And those of us that live next to the LHC wonder if we will be sucked into a {vortex|wormhole}. If you are, it won't matter if you live near it or not. Regards Marshall Thomas
Re: Superfast internet may replace world wide web
On Mon, 07 Apr 2008 17:36:09 +0200, Thomas Kernen said: And those of us that live next to the LHC wonder if we will be sucked into a {vortex|wormhole}. You mean like this? http://ars.userfriendly.org/cartoons/?id=20080406mode=classic pgplzlVbya2JN.pgp Description: PGP signature
RE: Superfast internet may replace world wide web
Subject: Superfast internet may replace world wide web says the solemn headline of Telegraph. Hasn't your mummy told you not to believe everything that you read in the papers? Especially when it involves technology! In any case, there is no new Internet here, just an engineered P2P network (or call it a CDN if you will) that is intended to distribute 15 million gigs per year of data to scientists who crunch that data on virtual supercomputer clusters known as the Grid. They do all of this on the Internet today, except for big data transfers for which most countries have build special academic IP networks. The Grid is rather like Amazon's EC2 and this CERN project is rather like Amazon's S3. Yes, I agree with the Telegraph that P2P and cloud computing Amazon style, are indeed the wave of the future, but they won't replace the web or the Internet. They are just another theme being added to the Internet recipe. It's just like Heston Blumenthal's cuisine http://en.wikipedia.org/wiki/Heston_Blumenthal; it's still food, it's still served in restaurants and it still counts towards his three Michelin stars. Still, I don't expect bacon and eggs ice cream to come to Baskin Robbins anytime soon. --Michael Dillon
Train wreck (was Does TCP Need an Overhaul?)
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Apr 7, 2008, at 8:36 AM, Lucy Lynch wrote: Anyone out there attend this event? The Future of TCP: Train-wreck or Evolution? http://yuba.stanford.edu/trainwreck/agenda.html how did the demos go? The researchers demonstrated four things that made sense to me: (1) TCP is not the right transport for carrying video data if what you want is real-time delivery. Carrying stored video (YouTube-style) is fine, but if you're trying to watch TV, you really should be using some other transport such as RTP or DCCP. Same comment holds for sensor traffic, but the astronomers who carry radiotelescope data halfway around the world weren't present. (2) TCP is probably not the right protocol for carrying transaction traffic within a data center. One speculates that SCTP (which has a concept of a stream of TCP-like transactions that can be handled out of order and allows for congestion management both within and among transactions) might be a better protocol, and in any event that when thousands of transactions back up in a gigabit Ethernet chip's queue on a host that the host should start noticing that they are experiencing congestion. (3) 802.11 networks experience not only the traditional congestion experienced in wired networks, but channel access congestion (true of shared media in general) and radio interference. In such networks, it may be useful to think about congestion as happening in a region as opposed to at a bottleneck. (4) When it is pointed out that instead of complaining about TCP in cases where it is the wrong protocol it may be more useful to use the transport designed for the purpose, researchers who presumably are expert on matters in the transport layer respond in complete surprise. -BEGIN PGP SIGNATURE- iD8DBQFH+klNbjEdbHIsm0MRAlLhAKCDprgXaKYukFG57KRsRS8HyGAUHgCgyRLd SpNahEUbZudgcoc3bMz/Cto= =hnGa -END PGP SIGNATURE-
Re: Any tool or theorical method on detecting number of computer behind a NAT box?
On Mon, 7 Apr 2008 23:51:55 +0800 (CST) Joe Shen [EMAIL PROTECTED] wrote: hi, Sharing internet access bandwidth between multiple computers is common today. Usually, bandwidth sharer bought a little router with NAT/PAT function. After connecting that box to a ADSL/LAN access link, multiple computer could share a single access link. I heard some company provide prdouct for detecting number of computers behind NAT/PAT box. Is there any paper or document on how such product work? where could I fint them ? THere was a paper of mine from a few years ago: http://www.cs.columbia.edu/~smb/papers/fnat.pdf --Steve Bellovin, http://www.cs.columbia.edu/~smb
Re: Does TCP Need an Overhaul? (internetevolution, via slashdot)
On 7 apr 2008, at 16:20, Kevin Day wrote: As a quick example on two FreeBSD 7.0 boxes attached directly over GigE, with New Reno, fast retransmit/recovery, and 256K window sizes, with an intermediary router simulating packet loss. A single HTTP TCP session going from a server to client. Ok, assuming a 1460 MSS that leaves the RTT as the unknown. SACK enabled, 0% packet loss: 780Mbps SACK disabled, 0% packet loss: 780Mbps Is that all? Try with jumboframes. SACK enabled, 0.005% packet loss: 734Mbps SACK disabled, 0.005% packet loss: 144Mbps (19.6% the speed of having SACK enabled) 144 Mbps and 0.5 packet loss probability would result in a ~ 110 ms RTT so obviously something isn't right with that case. 734 would be an RTT of around 2 ms, which sounds fairly reasonable. I'd be interested to see what's really going on here, I suspect that the packet loss isn't sufficiently random so multiple segments are lost from a single window. Or maybe disabling SACK also disables fast retransmit? I'll be happy to look at a tcpdump for the 144 Mbps case. It would be very nice if more network-friendly protocols were in use, but with download optimizers for Windows that cranks the TCP window sizes way up, the general move to solving latency by opening more sockets, and P2P doing whatever it can to evade ISP detection - it's probably a bit late. Don't forget that the user is only partially in control, the data also has to come from somewhere. Service operators have little incentive to break the network. And users would probably actually like it if their p2p was less aggressive, that way you can keep it running when you do other stuff without jumping through traffic limiting hoops.
Re: Train wreck (was Does TCP Need an Overhaul?)
On 7 apr 2008, at 18:18, Fred Baker wrote: (4) When it is pointed out that instead of complaining about TCP in cases where it is the wrong protocol it may be more useful to use the transport designed for the purpose, researchers who presumably are expert on matters in the transport layer respond in complete surprise. There is of course the issue of migrating from one transport to another with NATs and firewalls thrown in for good measure, which is worse than migrating to IPv6 in some ways and only significantly better in one (no need to upgrade routers).
Re: Superfast internet may replace world wide web
On Mon, 7 Apr 2008, Bill Woodcock wrote: On Mon, 7 Apr 2008, Glen Kent wrote: says the solemn headline of Telegraph. .. and we in Nanog are still discussing IPv6! ;-) It's because we don't have a hadron demolition derby to power our American interwebs: The power of the grid will be unlocked this summer with the switching on of the Large Hadron Collider (LHC). I doubt it: http://www.geant2.net/server/show/nav.00d00h001003 The structure of the LHC Computing Grid (LCG) is to distribute the data first to 12 Tier 1 sites, each connected to Tier 0 (CERN) by a dedicated wavelength switched path of 10Gbps. These paths are provided by the new hybrid (IP routed/ wavelength switched) structure of G?ANT2. Corresponding dark-fibre lightpaths will be provided by each of the European NRENs involved. So within Europe, much of the LHC data will move via paths that not even part of the Internet/Geant2 infrastructure: http://www.geant2.net/upload/pdf/PUB-07-179_GN2_Topology_Jan_08_final.pdf Just look for the black links which are all dark fiber out of Switzerland. Also, LHC will generate 15Petabytes per annum - not Gigabytes or Terabytes as some media have stated. -Hank
Re: Superfast internet may replace world wide web
On Apr 7, 2008, at 11:39 AM, Steven M. Bellovin wrote: On Mon, 7 Apr 2008 08:24:54 -0700 (PDT) Lucy Lynch [EMAIL PROTECTED] wrote: http://xkcd.com/401/ Also http://ars.userfriendly.org/cartoons/?id=20080330 and http://ars.userfriendly.org/cartoons/?id=20080406 I love those! I also love the top story here, especially the last sentence: http://bobpark.physics.umd.edu/WN08/wn040408.html -- TTFN, patrick
Re: Superfast internet may replace world wide web
Date: Mon, 7 Apr 2008 20:21:26 +0530 From: Glen Kent [EMAIL PROTECTED] Sender: [EMAIL PROTECTED] says the solemn headline of Telegraph. http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/06/ninternet106.xml Also related to this one, here: Web could collapse as video demand soars http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/07/nweb107.xml .. and we in Nanog are still discussing IPv6! ;-) Sigh. Never let a clueless writer put up a story as technically complex as this. He clearly does not know the difference between the web (which WAS invented at CERN) and the Internet (which was not). His confusion on this and other details leads to a story which has little or nothing to say. 1. The grid was NOT invented at CERN, although CERN/LHC people were involved. 2. Aside from being the a huge physics experiment, it is also a huge network experiment. We will be carrying many gigabits of data from CERN to FermiLab and Brookhaven as well as from those facilities to physics researcher all over the world. By 2011 we may be seeing close to 100 Gbps 24/7 for months at a time. And that is just data from CERN to the US. They will be sending data to many other countries. (OK, there are some short pauses for calibration.) 3. This will all be over the Internet, though much will utilize dedicated lines purchased/leased just for this. But it's still TCP/IP and UDP (mostly the former) and mostly using traditional P2P techniques to get adequate performance over links with RTTs in excess of 200 ms. It is true that the problems faced by CERN are similar to those faced by CDNs streaming video, but it is different in that this data is NOT streamed. You can't take the chance that the packet with the Higgs Boson waving hello is dropped. Since almost of the traffic is passing over dedicated links, congestion due to aggregation, the big issue with streaming video, is simply not an issue. We want to move as much data in a single stream as you can convince TCP to allow. So the things learned from the LHC network experiment may well help improve the Internet and help with things like video distribution, the grid is NOT going to replace the web, let alone the Internet. -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: [EMAIL PROTECTED] Phone: +1 510 486-8634 Key fingerprint:059B 2DDF 031C 9BA3 14A4 EADA 927D EBB3 987B 3751 pgpyfWrw6l2qz.pgp Description: PGP signature
Re: Superfast internet may replace world wide web
On Mon, 07 Apr 2008 20:21:26 +0530, Glen Kent said: says the solemn headline of Telegraph. http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/06/ninternet106.xml So yoy get higher bandwidth (physical pipe allowing) by downloading from a grid of systems. Sounds suspiciously like somebody has re-invented BitTorrent? (Sorry, am in a cynical mood today.. ;) pgpa8tbzcFIzG.pgp Description: PGP signature
Re: Superfast internet may replace world wide web
That and someone can't tell the difference between a network and an application that runs in a network. On Apr 7, 2008, at 10:38 AM, [EMAIL PROTECTED] wrote: On Mon, 07 Apr 2008 20:21:26 +0530, Glen Kent said: says the solemn headline of Telegraph. http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/06/ ninternet106.xml So yoy get higher bandwidth (physical pipe allowing) by downloading from a grid of systems. Sounds suspiciously like somebody has re-invented BitTorrent? (Sorry, am in a cynical mood today.. ;)
Re: Superfast internet may replace world wide web
On Apr 7, 2008, at 1:00 PM, Kevin Oberman wrote: Date: Mon, 7 Apr 2008 20:21:26 +0530 From: Glen Kent [EMAIL PROTECTED] Sender: [EMAIL PROTECTED] says the solemn headline of Telegraph. http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/06/ninternet106.xml Also related to this one, here: Web could collapse as video demand soars http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/07/nweb107.xml .. and we in Nanog are still discussing IPv6! ;-) Sigh. Never let a clueless writer put up a story as technically complex as this. He clearly does not know the difference between the web (which WAS invented at CERN) and the Internet (which was not). His confusion on this and other details leads to a story which has little or nothing to say. 1. The grid was NOT invented at CERN, although CERN/LHC people were involved. 2. Aside from being the a huge physics experiment, it is also a huge network experiment. We will be carrying many gigabits of data from CERN to FermiLab and Brookhaven as well as from those facilities to physics researcher all over the world. By 2011 we may be seeing close to 100 Gbps 24/7 for months at a time. And that is just data from CERN to the US. They will be sending data to many other countries. (OK, there are some short pauses for calibration.) 3. This will all be over the Internet, though much will utilize dedicated lines purchased/leased just for this. But it's still TCP/IP and UDP (mostly the former) and mostly using traditional P2P techniques to get adequate performance over links with RTTs in excess of 200 ms. It is true that the problems faced by CERN are similar to those faced by CDNs streaming video, but it is different in that this data is NOT streamed. You can't take the chance that the packet with the Higgs Boson waving hello is dropped. I would actually disagree with that, _IF_ your SNR is limited by your bit rate. In VLBI (where the SNR _IS_ limited by the bit rate) it is more efficient to send more (new) data than to repeat old data that gets lost. Having talked to particle physicists here who feel that they are in the same regime, I would be curious as to whether or not CERN has done the math on this, and with what result. Regards Marshall Since almost of the traffic is passing over dedicated links, congestion due to aggregation, the big issue with streaming video, is simply not an issue. We want to move as much data in a single stream as you can convince TCP to allow. So the things learned from the LHC network experiment may well help improve the Internet and help with things like video distribution, the grid is NOT going to replace the web, let alone the Internet. -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: [EMAIL PROTECTED] Phone: +1 510 486-8634 Key fingerprint:059B 2DDF 031C 9BA3 14A4 EADA 927D EBB3 987B 3751
Bandwidth issues in the Sprint network
I am currently having problems get upload bandwidth on a Sprint circuit. I am using a full OC3 circuit. I am doing fine on downloading data, but uploading data I can only get about 5Mbps with ftp or a speedtest. I have tested against multiple networks and this has stayed the same. Monitoring Cacti graphs and the router I do get about 30Mbps total traffic outbound, but individual (flows/ip?) test always seem limited. I would like to know if anyone else sees anything similar, or where I can get help. The assistance I have gotten from Sprint up to this point is that they find no problems. Due to the consistency of 5Mbps I am suspecting rate limiting, but wanted to know if I was overlooking something else. -- Brian Raaen Network Engineer [EMAIL PROTECTED] signature.asc Description: This is a digitally signed message part.
Re: Bandwidth issues in the Sprint network
On Mon, 07 Apr 2008 15:06:21 EDT, Brian Raaen said: have gotten from Sprint up to this point is that they find no problems. Due to the consistency of 5Mbps I am suspecting rate limiting, but wanted to know if I was overlooking something else. TCP window size tuning? I'd look there first... pgp5sXCAaiPAS.pgp Description: PGP signature
Perspectives: Diseconomies of Scale [was: Re: (latency) cooling door ...
I offer this as an afterthought without any expectations of replies to last week's latency discussion that focused on centralized vs. distributed, or one vs. many, data centers. While IMO the authors do not take into proper account the costs associated with network bandwidth and node provisioning, I thought it was nonetheless interesting due to the other factors they highlighted. There's nothing particularly correct here, merely some additional viewpoints on the subject that were not covered here earlier. Enjoy! Diseconomies of Scale - by Ken Church James Hamilton April 6, 2008 http://perspectives.mvdirona.com/2008/04/06/DiseconomiesOfScale.aspx Frank ---
Re: Bandwidth issues in the Sprint network
--- [EMAIL PROTECTED] wrote: I am currently having problems get upload bandwidth on a Sprint circuit. I am using a full OC3 circuit. I am doing fine on downloading data, but uploading data I can only get about 5Mbps with ftp or a speedtest. I have tested against multiple networks and this has stayed the same. Monitoring Cacti graphs and the router I do get about 30Mbps total traffic outbound, but individual (flows/ip?) test always seem limited. I would like to know if anyone else sees anything similar, or where I can get help. The assistance I have gotten from Sprint up to this point is that they find no problems. Due to the consistency of 5Mbps I am suspecting rate limiting, but wanted to know if I was overlooking something else. I would not use one FTP session to test bandwidth. The rate limiting may be in the FTP software or other area of the computer. Likewise, Speedtest servers (in my opinion) are more marketing tools than testing tools. Try several similarly configured (but separate boxes) FTP servers simultaneously. If you see it go up by a factor of three you've found the issue. I have had to push my four OC-12s to Sprint to the max at times and get full BW. That's in Hawaii, but I imagine it's the same as other areas. scott
RE: Bandwidth issues in the Sprint network
See if you can find a nother connector that can help with using iperf. Also, make sure any system testing systems have tuned IP stacks. That info is also linked from the iperf web page. http://dast.nlanr.net/Projects/Iperf/ http://www.psc.edu/networking/projects/tcptune/ Robert D. Scott [EMAIL PROTECTED] Senior Network Engineer 352-273-0113 Phone CNS - Network Services 352-392-2061 CNS Receptionist University of Florida 352-392-9440 FAX Florida Lambda Rail 352-294-3571 FLR NOC Gainesville, FL 32611 -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Scott Weeks Sent: Monday, April 07, 2008 5:24 PM To: nanog@merit.edu Subject: Re: Bandwidth issues in the Sprint network --- [EMAIL PROTECTED] wrote: I am currently having problems get upload bandwidth on a Sprint circuit. I am using a full OC3 circuit. I am doing fine on downloading data, but uploading data I can only get about 5Mbps with ftp or a speedtest. I have tested against multiple networks and this has stayed the same. Monitoring Cacti graphs and the router I do get about 30Mbps total traffic outbound, but individual (flows/ip?) test always seem limited. I would like to know if anyone else sees anything similar, or where I can get help. The assistance I have gotten from Sprint up to this point is that they find no problems. Due to the consistency of 5Mbps I am suspecting rate limiting, but wanted to know if I was overlooking something else. I would not use one FTP session to test bandwidth. The rate limiting may be in the FTP software or other area of the computer. Likewise, Speedtest servers (in my opinion) are more marketing tools than testing tools. Try several similarly configured (but separate boxes) FTP servers simultaneously. If you see it go up by a factor of three you've found the issue. I have had to push my four OC-12s to Sprint to the max at times and get full BW. That's in Hawaii, but I imagine it's the same as other areas. scott
Re: Bandwidth issues in the Sprint network
--- [EMAIL PROTECTED] wrote: --- I would like to second the recommendation and go one further. Internet2 has released a performance toolkit that is run from CD. I would like to -- Robert D. Scott wrote: See if you can find a nother connector that can help with using iperf. Also, --- The thing to note about most tools like these is you need a box on both sides of the circuit using the same software. One could be 'out there' on the internet, but the further 'out there' your other box is, the less valid your test is. scott
Iperf 2.0.4 Released
On the subject of iperf, I just received this today: Iperf 2.0.4 addresses one major and several minor issues with Iperf. The bugs fixed were: * Iperf should no longer consume gratuitous CPU under Linux The help messages * have been expanded to include previously undocumented options The header * the stats report was missing a header which has been replaced. New in Iperf 2.0.4: * Under Linux you can select the TCP congestion algorithm by using the -Z (--linux-congestion) flag * Iperf has a minimal man page! Thanks to Stephen Hemminger and Claus Klein for their patches. This is intended to be the last release in the 2.0 train. Development efforts going forward will concentrate on the 2.1 series of releases. If significant bugs are found there will be a 2.0.5 release, but hopefully we won't need to do that. You can download it at: https://sourceforge.net/project/showfiles.php?group_id=128336 -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: [EMAIL PROTECTED] Phone: +1 510 486-8634 Key fingerprint:059B 2DDF 031C 9BA3 14A4 EADA 927D EBB3 987B 3751 pgpYP3q8gyzYQ.pgp Description: PGP signature